Sparse Possibilistic Clustering with L1 Regularization
Ryo Inokuchi, Sadaaki Miyamoto · 2007 IEEE International Conference on Granular Computing (GRC 2007) · 2007
Possibilistic clustering is an efficient method to detect high density regions and more robust than fuzzy c-means. However, it is not 'sparse', since a cluster center is expressed as a linear combination of all data. In this paper, we propose a sparse possibilistic clustering method with 11 regularization to find compact clusters. Due to a non- negative constraints for a membership, the baseline constant is introduced into the regularizer. The effectiveness of the proposed method is shown in illustrative examples.